Estimating causal effects with the neural autoregressive density estimator

Abstract The estimation of causal effects is fundamental in situations where the underlying system will be subject to active interventions. Part of building a causal inference engine is defining how variables relate to each other, that is, defining the functional relationship between variables entailed by the graph conditional dependencies. In this article, we deviate from the common assumption of linear relationships in causal models by making use of neural autoregressive density estimators and use them to estimate causal effects within Pearl’s do-calculus framework. Using synthetic data, we show that the approach can retrieve causal effects from non-linear systems without explicitly modeling the interactions between the variables and include confidence bands using the non-parametric bootstrap. We also explore scenarios that deviate from the ideal causal effect estimation setting such as poor data support or unobserved confounders.

Download Full-text

Counterfactual Reasoning in Observational Studies

Proceedings of the AAAI Conference on Artificial Intelligence ◽

10.1609/aaai.v33i01.33019886 ◽

2019 ◽

Vol 33 ◽

pp. 9886-9887

Author(s):

Negar Hassanpour

Keyword(s):

Alternative Treatment ◽

Observational Studies ◽

Causal Effect ◽

Intelligent Agent ◽

Population Based ◽

Causal Effects ◽

Medical Procedure ◽

Counterfactual Reasoning ◽

Effect Estimation ◽

The Individual

To identify the appropriate action to take, an intelligent agent must infer the causal effects of every possible action choices. A prominent example is precision medicine that attempts to identify which medical procedure will benefit each individual patient the most. This requires answering counterfactual questions such as: ""Would this patient have lived longer, had she received an alternative treatment?"". In my PhD, I attempt to explore ways to address the challenges associated with causal effect estimation; with a focus on devising methods that enhance performance according to the individual-based measures (as opposed to population-based measures).

Download Full-text

Instrumental Heterogeneity in Sex-Specific Two-Sample Mendelian Randomization: Empirical Results From the Relationship Between Anthropometric Traits and Breast/Prostate Cancer

Frontiers in Genetics ◽

10.3389/fgene.2021.651332 ◽

2021 ◽

Vol 12 ◽

Author(s):

Yixin Gao ◽

Jinhui Zhang ◽

Huashuo Zhao ◽

Fengjun Guan ◽

Ping Zeng

Keyword(s):

Breast Cancer ◽

Prostate Cancer ◽

Mendelian Randomization ◽

Causal Effect ◽

Correlation Coefficients ◽

Causal Effects ◽

Mr Studies ◽

Pearson’S Correlation ◽

Pearson's Correlation ◽

Effect Estimation

BackgroundIn two-sample Mendelian randomization (MR) studies, sex instrumental heterogeneity is an important problem needed to address carefully, which however is often overlooked and may lead to misleading causal inference.MethodsWe first employed cross-trait linkage disequilibrium score regression (LDSC), Pearson’s correlation analysis, and the Cochran’s Q test to examine sex genetic similarity and heterogeneity in instrumental variables (IVs) of exposures. Simulation was further performed to explore the influence of sex instrumental heterogeneity on causal effect estimation in sex-specific two-sample MR analyses. Furthermore, we chose breast/prostate cancer as outcome and four anthropometric traits as exposures as an illustrative example to illustrate the importance of taking sex heterogeneity of instruments into account in MR studies.ResultsThe simulation definitively demonstrated that sex-combined IVs can lead to biased causal effect estimates in sex-specific two-sample MR studies. In our real applications, both LDSC and Pearson’s correlation analyses showed high genetic correlation between sex-combined and sex-specific IVs of the four anthropometric traits, while nearly all the correlation coefficients were larger than zero but less than one. The Cochran’s Q test also displayed sex heterogeneity for some instruments. When applying sex-specific instruments, significant discrepancies in the magnitude of estimated causal effects were detected for body mass index (BMI) on breast cancer (P = 1.63E-6), for hip circumference (HIP) on breast cancer (P = 1.25E-20), and for waist circumference (WC) on prostate cancer (P = 0.007) compared with those generated with sex-combined instruments.ConclusionOur study reveals that the sex instrumental heterogeneity has non-ignorable impact on sex-specific two-sample MR studies and the causal effects of anthropometric traits on breast/prostate cancer would be biased if sex-combined IVs are incorrectly employed.

Download Full-text

Causal effects and counterfactual conditionals: contrasting Rubin, Lewis and Pearl

Economics and Philosophy ◽

10.1017/s0266267120000437 ◽

2021 ◽

pp. 1-21

Author(s):

Keith A. Markus

Keyword(s):

Comparative Research ◽

Causal Effect ◽

Weak Form ◽

Causal Effects ◽

Rival Theory ◽

Counterfactual Conditionals ◽

Effect Estimation ◽

The Many ◽

Rule Out

Abstract Rubin and Pearl offered approaches to causal effect estimation and Lewis and Pearl offered theories of counterfactual conditionals. Arguments offered by Pearl and his collaborators support a weak form of equivalence such that notation from the rival theory can be re-purposed to express Pearl’s theory in a way that is equivalent to Pearl’s theory expressed in its native notation. Nonetheless, the many fundamental differences between the theories rule out any stronger form of equivalence. A renewed emphasis on comparative research can help to guide applications, further develop each theory, and better understand their relative strengths and weaknesses.

Download Full-text

Multi-Cause Effect Estimation with Disentangled Confounder Representation

Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence ◽

10.24963/ijcai.2021/384 ◽

2021 ◽

Author(s):

Jing Ma ◽

Ruocheng Guo ◽

Aidong Zhang ◽

Jundong Li

Keyword(s):

Real World ◽

Fundamental Problem ◽

Causal Effect ◽

Causal Effects ◽

Treatment Effect Estimation ◽

Latent Representations ◽

Real World Datasets ◽

New Perspective ◽

Multiple Treatments ◽

Effect Estimation

One fundamental problem in causality learning is to estimate the causal effects of one or multiple treatments (e.g., medicines in the prescription) on an important outcome (e.g., cure of a disease). One major challenge of causal effect estimation is the existence of unobserved confounders -- the unobserved variables that affect both the treatments and the outcome. Recent studies have shown that by modeling how instances are assigned with different treatments together, the patterns of unobserved confounders can be captured through their learned latent representations. However, the interpretability of the representations in these works is limited. In this paper, we focus on the multi-cause effect estimation problem from a new perspective by learning disentangled representations of confounders. The disentangled representations not only facilitate the treatment effect estimation but also strengthen the understanding of causality learning process. Experimental results on both synthetic and real-world datasets show the superiority of our proposed framework from different aspects.

Download Full-text

A Bayesian Approach for Estimating Causal Effects from Observational Data

Proceedings of the AAAI Conference on Artificial Intelligence ◽

10.1609/aaai.v34i04.5988 ◽

2020 ◽

Vol 34 (04) ◽

pp. 5395-5402

Author(s):

Johan Pensar ◽

Topi Talvitie ◽

Antti Hyttinen ◽

Mikko Koivisto

Keyword(s):

Bayesian Method ◽

Causal Relation ◽

Causal Effect ◽

Small Sample ◽

Causal Effects ◽

Estimation Accuracy ◽

Flow Cytometry Data ◽

Mere Existence ◽

Effect Estimation ◽

Small Sample Sizes

We present a novel Bayesian method for the challenging task of estimating causal effects from passively observed data when the underlying causal DAG structure is unknown. To rigorously capture the inherent uncertainty associated with the estimate, our method builds a Bayesian posterior distribution of the linear causal effect, by integrating Bayesian linear regression and averaging over DAGs. For computing the exact posterior for all cause-effect variable pairs, we give an algorithm that runs in time O(3d d) for d variables, being feasible up to 20 variables. We also give a variant that computes the posterior probabilities of all pairwise ancestor relations within the same time complexity, significantly improving the fastest previous algorithm. In simulations, our Bayesian method outperforms previous methods in estimation accuracy, especially for small sample sizes. We further show that our method for effect estimation is well-adapted for detecting strong causal effects markedly deviating from zero, while our variant for computing posteriors of ancestor relations is the method of choice for detecting the mere existence of a causal relation. Finally, we apply our method on observational flow cytometry data, detecting several causal relations that concur with previous findings from experimental data.

Download Full-text

Student Perceptions of Community: A Different Perspective

NASPA Journal ◽

10.2202/0027-6014.1064 ◽

1998 ◽

Vol 35 (4) ◽

Author(s):

Jackie Clark ◽

Joan Hirt

Keyword(s):

Student Perceptions ◽

Large Scale ◽

Research University ◽

Small Scale ◽

Social Environments ◽

Common Assumption ◽

Small Communities ◽

Public Research University ◽

The Common ◽

Using Data

The creation of small communities has been proposed as a way of enhancing the educational experience of students at large institutions. Using data from a survey of students living in large and small residences at a public research university, this study does not support the common assumption that small-scale social environments are more conducive to positive community life than large-scale social environments.

Download Full-text

Understanding the nature of association between anxiety phenotypes and anorexia nervosa: a triangulation approach

BMC Psychiatry ◽

10.1186/s12888-020-02883-8 ◽

2020 ◽

Vol 20 (1) ◽

Author(s):

E. Caitlin Lloyd ◽

Hannah M. Sallis ◽

Bas Verplanken ◽

Anne M. Haase ◽

Marcus R. Munafò

Keyword(s):

Anorexia Nervosa ◽

Longitudinal Study ◽

Anxiety Disorder ◽

Anxiety Disorders ◽

Causal Effect ◽

Causal Effects ◽

Observational Research ◽

Genome Wide Association Studies ◽

Genetic Liability ◽

Depressed Affect

Abstract Background Evidence from observational studies suggests an association between anxiety disorders and anorexia nervosa (AN), but causal inference is complicated by the potential for confounding in these studies. We triangulate evidence across a longitudinal study and a Mendelian randomization (MR) study, to evaluate whether there is support for anxiety disorder phenotypes exerting a causal effect on AN risk. Methods Study One assessed longitudinal associations of childhood worry and anxiety disorders with lifetime AN in the Avon Longitudinal Study of Parents and Children cohort. Study Two used two-sample MR to evaluate: causal effects of worry, and genetic liability to anxiety disorders, on AN risk; causal effects of genetic liability to AN on anxiety outcomes; and the causal influence of worry on anxiety disorder development. The independence of effects of worry, relative to depressed affect, on AN and anxiety disorder outcomes, was explored using multivariable MR. Analyses were completed using summary statistics from recent genome-wide association studies. Results Study One did not support an association between worry and subsequent AN, but there was strong evidence for anxiety disorders predicting increased risk of AN. Study Two outcomes supported worry causally increasing AN risk, but did not support a causal effect of anxiety disorders on AN development, or of AN on anxiety disorders/worry. Findings also indicated that worry causally influences anxiety disorder development. Multivariable analysis estimates suggested the influence of worry on both AN and anxiety disorders was independent of depressed affect. Conclusions Overall our results provide mixed evidence regarding the causal role of anxiety exposures in AN aetiology. The inconsistency between outcomes of Studies One and Two may be explained by limitations surrounding worry assessment in Study One, confounding of the anxiety disorder and AN association in observational research, and low power in MR analyses probing causal effects of genetic liability to anxiety disorders. The evidence for worry acting as a causal risk factor for anxiety disorders and AN supports targeting worry for prevention of both outcomes. Further research should clarify how a tendency to worry translates into AN risk, and whether anxiety disorder pathology exerts any causal effect on AN.

Download Full-text

Lack of an association between gallstone disease and bilirubin levels with risk of colorectal cancer: a Mendelian randomisation analysis

British Journal of Cancer ◽

10.1038/s41416-020-01211-x ◽

2021 ◽

Author(s):

Richard Culliford ◽

Alex J. Cornish ◽

Philip J. Law ◽

Susan M. Farrington ◽

Kimmo Palin ◽

...

Keyword(s):

Colorectal Cancer ◽

Large Scale ◽

Causal Effect ◽

Gallstone Disease ◽

Epidemiological Studies ◽

Mendelian Randomisation ◽

Linear Relationships ◽

Potential Risk Factors ◽

The Relationship ◽

Circulating Levels

Abstract Background Epidemiological studies of the relationship between gallstone disease and circulating levels of bilirubin with risk of developing colorectal cancer (CRC) have been inconsistent. To address possible confounding and reverse causation, we examine the relationship between these potential risk factors and CRC using Mendelian randomisation (MR). Methods We used two-sample MR to examine the relationship between genetic liability to gallstone disease and circulating levels of bilirubin with CRC in 26,397 patients and 41,481 controls. We calculated the odds ratio per genetically predicted SD unit increase in log bilirubin levels (ORSD) for CRC and tested for a non-zero causal effect of gallstones on CRC. Sensitivity analysis was applied to identify violations of estimator assumptions. Results No association between either gallstone disease (P value = 0.60) or circulating levels of bilirubin (ORSD = 1.00, 95% confidence interval (CI) = 0.96–1.03, P value = 0.90) with CRC was shown. Conclusions Despite the large scale of this study, we found no evidence for a causal relationship between either circulating levels of bilirubin or gallstone disease with risk of developing CRC. While the magnitude of effect suggested by some observational studies can confidently be excluded, we cannot exclude the possibility of smaller effect sizes and non-linear relationships.

Download Full-text

Śaivism after the Śaiva Age: Continuities in the Scriptural Corpus of the Vīramāheśvaras

Religions ◽

10.3390/rel12030222 ◽

2021 ◽

Vol 12 (3) ◽

pp. 222

Author(s):

Elaine M. Fisher

Keyword(s):

Thirteenth Century ◽

Common Assumption ◽

Academic Practice ◽

The Common ◽

Popular Work ◽

First Time

This article makes the case that Vīraśaivism emerged in direct textual continuity with the tantric traditions of the Śaiva Age. In academic practice up through the present day, the study of Śaivism, through Sanskrit sources, and bhakti Hinduism, through the vernacular, are generally treated as distinct disciplines and objects of study. As a result, Vīraśaivism has yet to be systematically approached through a philological analysis of its precursors from earlier Śaiva traditions. With this aim in mind, I begin by documenting for the first time that a thirteenth-century Sanskrit work of what I have called the Vīramāheśvara textual corpus, the Somanāthabhāṣya or Vīramāheśvarācārasāroddhārabhāṣya, was most likely authored by Pālkurikĕ Somanātha, best known for his vernacular Telugu Vīraśaiva literature. Second, I outline the indebtedness of the early Sanskrit and Telugu Vīramāheśvara corpus to a popular work of early lay Śaivism, the Śivadharmaśāstra, with particular attention to the concepts of the jaṅgama and the iṣṭaliṅga. That the Vīramāheśvaras borrowed many of their formative concepts and practices directly from the Śivadharmaśāstra and other works of the Śaiva Age, I argue, belies the common assumption that Vīraśaivism originated as a social and religious revolution.

Download Full-text

The Pice Effects of Competition from Parallel Imports and Therapeutic Alternatives: Using Dynamic Models to Estimate the Causal Effect on the Extensive and Intensive Margins

Review of Industrial Organization ◽

10.1007/s11151-021-09834-x ◽

2021 ◽

Author(s):

David Granlund

Keyword(s):

Dynamic Models ◽

Causal Effect ◽

The Other ◽

Causal Effects ◽

Parallel Imports ◽

Extensive And Intensive Margins ◽

Price Effects ◽

Short And Long Term ◽

Therapeutic Alternatives

AbstractThis paper studies responses to competition with the use of dynamic models that distinguish between short- and long-term price effects. The dynamic models also allow lagged numbers of competitors to become valid and strong instruments for the current numbers, which enables studying the causal effects using flexible specifications. A first parallel trader is found to decrease prices of exchangeable products by 7% in the long term. On the other hand, prices do not respond to the first competitor that sells therapeutic alternatives; but competition from four or more competitors that sell on-patent therapeutic alternatives decreases prices by about 10% in the long term.

Download Full-text